betaFeedback

WorksPhysics and Astronomy

Direct numerical simulations of turbulent mixing driven by the Faraday instability in rotating miscible fluids

Abstract

The effect of the rotation on the turbulent mixing of two miscible fluids of small contrasting density, induced by Faraday instability, is investigated using direct numerical simulations (DNS). We quantify the irreversible mixing which depicts the conversion of the available potential energy (APE) to the background potential energy (BPE) through irreversible mixing rate M\mathcal{M}. We demonstrate that at lower forcing amplitudes, the turbulent kinetic energy (t.k.e.t.k.e.) increases with an increase in the Coriolis frequency ff till (f/ω)2<0.25\left(f/ω\right)^2<0.25, where ωω is the forcing frequency, during the sub-harmonic instability phase. This enhancement of t.k.e.t.k.e. is attributed to the excitement of more unstable modes. The irreversible mixing sustains for an extended period with increasing (f/ω)2\left(f/ω\right)^2 till 0.250.25 owing to the prolonged sub-harmonic instability phase and eventually ceases with instability saturation. When (f/ω)2>0.25\left(f/ω\right)^2 > 0.25, the Coriolis force significantly delays the onset of the sub-harmonic instabilities. The strong rotational effects result in lower turbulence because the bulk of the APE expends to BPE, decreasing APE that converts back to t.k.e.t.k.e. reservoir for (f/ω)2>0.25\left(f/ω\right)^2 > 0.25. Since the instability never saturates for (f/ω)2>0.25\left(f/ω\right)^2 > 0.25, conversion of APE to BPE via M\mathcal{M} continues, and we find prolonged irreversible mixing. At higher forcing amplitudes, the instability delaying effect of rotation is negligible, and the turbulence is less intense and short-lived. Therefore, the irreversible mixing phenomenon also ends quickly for (f/ω)2<0.25\left(f/ω\right)^2<0.25. However, when (f/ω)2>0.25\left(f/ω\right)^2>0.25, a continuous irreversible mixing is observed.

Banner appropriate to article type will appear here in typeset article

† Email address for correspondence: [email protected]

Introduction

The Faraday instability (Faraday 1831) describes the generation of standing wave patterns at the interface of two immiscible fluids subjected to vertical periodic vibration. However, in a stably stratified two-layer miscible fluid system, these standing waves become highly disorganized above a certain forcing amplitude and start to interact with each other resulting in the mixing of fluids which eventually saturates [22] (Amiroudine et al. 2012; Diwakar et al. 2015). Gréa & Adou (2018); Briard et al. (2019) studied the onset and saturation of turbulent mixing owing to Faraday instability in fluids of small contrasting density for a wide range of parameters using theoretical models and DNS. They demonstrated that the diffuse interface begins to oscillate with small amplitude in the harmonic instability phase, followed by the rapid growth and development of a turbulent mixing zone in the sub-harmonic instability phase. However, owing to the saturation of the instability, the turbulent mixing cannot be sustained for long. They also estimated the final size of the mixing zone analytically as follows:

Lsat=2Ag0ω2(2F+4),(1)L_{sat} = \frac{2\mathcal{A}g_0}{\omega^2}(2F+4), \tag*{(1)}

where LsatL_{sat} is the final asymptotic size of the mixing zone (defined later) in saturation state, A=(ρ1−ρ2)/(ρ1+ρ2)\mathcal{A} = (\rho_1-\rho_2)/(\rho_1+\rho_2) is the Atwood number expressing the density contrast between heavy fluid (ρ1\rho_1) and light fluid (ρ2\rho_2), g0g_0 is the mean acceleration, FF is the oscillation amplitude and ω\omega is the forcing frequency. Briard et al. (2019) numerically presented the quantification of irreversible mixing associated with turbulent mixing by partitioning the total potential energy (PE) into background potential energy (BPE) and available potential energy (APE). They reported that the BPE signifying the measure of irreversible mixing increases during the onset of sub-harmonic instability. The APE, which represents the fraction of the total PE that can be transferred to BPE through irreversible mixing, peaks at saturation and is partially released in the flow as BPE. This results in more irreversible mixing, which causes the numerically computed final mixing zone size LL to exceed the analytically predicted LsatL_{sat} (1). Briard et al. (2020) performed a linear stability analysis accounting for the vertical inhomogeneity of the background density profile and viscous dissipation to predict the onset of the Faraday instability, either from a sharp or a diffuse interface and its eventual saturated state. Further experiments were carried out by Briard et al. (2020); Cavelier et al. (2022) using fresh and salty water to explore the dynamics of turbulent mixing, and DNS and theoretical predictions confirmed their findings. They found that when the instability is triggered, a natural wavelength appears at the interface between the two fluids. With the increase in amplitude, well-defined structures form that break due to the destabilization process at the nodes to produce turbulent mixing. Turbulence is eventually inhibited by instability saturation, and the final size of the mixing layer is consistent with the analytically predicted LsatL_{sat}. Recently, Singh & Pal (2022) presented a theoretical analysis of Faraday instability in miscible fluids under the effect of rotation and corroborated their findings with DNS. They reported that rotational effects delay the onset of the sub-harmonic instability at lower forcing amplitudes (FF), owing to the presence of a stable region between the harmonic to sub-harmonic transition and before the onset of the sub-harmonic instabilities. In the stable region, the mixing layer grows due to diffusion. However, with an increase in FF, the instability delaying effect of rotation diminishes due to the increasing area of all the unstable regions in the 3D stability diagram, resulting in the majority of excited modes being unstable. They concluded that the instability saturates for (f/ω)2<0.25(f/\omega)^2 < 0.25 and the mixing zone size asymptotes. However, for (f/ω)2≥0.25(f/\omega)^2 \ge0.25, the instability does not saturate, and the mixing zone size continues to grow.

These studies have demonstrated the potential of the Faraday instability to promote mixing in stably stratified fluids depending on the various parameters, such as frequency and amplitude of the vibrations and the strength of the initial stratification. Understanding the turbulent diapycnal mixing in stably stratified ocean interior is important as it plays a major role in the global overturning circulation that transports water mass, heat, nutrients, and other biochemical substances [21, 18]. The Faraday instability provides a framework to understand these flows and can help to elucidate the underlying physical processes that govern mixing in the oceanographic flows. The Earth’s rotation significantly affects the dynamics of diapycnal mixing processes in the thermocline, as reported in studies by Noh & Long [10]; Fernando [2] and Fleury et al. [3]. For example, N/f≈4.67N/f \approx4.67, where NN is the Brunt Väisälä frequency, in the abyssal southern ocean at mid-latitude [9, 16]. Modeling the properties of turbulence and the associated irreversible mixing, where external forcing, Earth’s rotation and stratification play a central role, is a major challenge in geophysical fluid dynamics. Motivated by the turbulent mixing in the oceanic flows, we aim to quantify the irreversible mixing and mixing efficiencies associated with the turbulent mixing driven by the time-periodic accelerations in rotating miscible fluids using DNS. In the present investigations, we follow the framework of Winters et al. [20], which was used to study the energetics of shear-induced mixing of stably stratified fluids by partitioning the total PE into BPE and APE. It is important to note that the Faraday instability drives the turbulent mixing, which induces small-scale turbulence and vertical motions that can enhance the mixing of stably stratified fluids. However, the turbulent mixing in stratified shear flows is typically driven by the Kelvin-Helmholtz instability resulting from velocity shear across the density gradient. As a result, the dynamics of the mixing layer differ between these two scenarios. Here, we focus on how the rotation affects the formation and subsequent breakdown of the structures on the diffuse interface to produce turbulence, irreversible mixing, and mixing efficiencies. To address these aims, the paper is organized as follows. In § 2, we discuss the governing equations with boundary equations and the set-up for the simulations and briefly explain the theoretical framework for linear stability analysis developed in Singh & Pal [17]. We then discuss the predictions of the linear stability analysis and the numerical results in § 3. Finally, we conclude in § 4.

Methods

Governing equations

The two-layer miscible fluid system with small density contrast is considered here, which is driven by vertical periodic oscillations of acceleration g(t)=g0(1+Fcos⁡(ωt))g(t) = g_0(1+F\cos(\omega t)). We consider the density of the mixture to be a linearly varying function of mass concentration such that the lighter fluid with concentration C(ρ2)=0C(\rho_2)=0 is placed above the denser fluid with C(ρ1)=1C(\rho_1)=1 in a rectangular domain. The unsteady incompressible flow with Boussinesq approximation is governed by the three-dimensional conservation equations for mass, momentum, and concentration field under vertically periodic forcing g(t)g(t), in the form [17]

∇⋅U=0,(2)\nabla\cdot U = 0, \tag*{(2)}
∂U∂t+U⋅∇U=−∇P−fk^×U+ν∇2U−2ACg(t)k^,(3)\frac{\partial U}{\partial t} + U \cdot\nabla U = -\nabla P - f\hat{k} \times U + \nu\nabla^2 U - 2\mathcal{A}C g(t)\hat{k}, \tag*{(3)}
∂C∂t+U⋅∇C=κ∇2C.(4)\frac{\partial C}{\partial t} + U \cdot\nabla C = \kappa\nabla^2 C. \tag*{(4)}

Here, UU denotes the velocity vector with components (U1,U2,U3)(U_1, U_2, U_3) in the horizontal x1x_1, x2x_2 and vertical x3x_3 directions respectively, PP is the pressure deviation from the hydrostatic background state, ν\nu (m2^2s$^{-1}) is the kinematic viscosity, and κ\kappa (m2^2s$^{-1}) is the diffusion coefficient. The Reynolds decomposition is performed to decompose a variable, say BB, into its horizontally averaged mean component ⟨B⟩H\langle B\rangle_H and fluctuating component bb, given by

B(x1,x2,x3,t)=⟨B⟩H(x3,t)+b(x1,x2,x3,t),(5)B(x_1,x_2,x_3,t)=\langle B\rangle_H(x_3,t)+b(x_1,x_2,x_3,t), \tag*{(5)}
⟨B⟩H(x3,t)=1lx1lx2∫−lx1/2+lx1/2∫−lx2/2+lx2/2B(x1,x2,x3,t) dx1dx2.(6)\langle B\rangle_H(x_3,t)=\frac{1}{l_{x_1}l_{x_2}}\int_{-l_{x_1}/2}^{+l_{x_1}/2}\int_{-l_{x_2}/2}^{+l_{x_2}/2}B(x_1,x_2,x_3,t)\,\mathrm{d}x_1\mathrm{d}x_2. \tag*{(6)}

Here, ⟨ ⟩H\langle\ \rangle_H denotes the horizontal average in the homogeneous directions x1x_1 and x2x_2, ⟨b⟩H=0\langle b\rangle_H=0, and lx1l_{x_1}, lx2l_{x_2} are the domain lengths in the x1x_1 and x2x_2 directions. The mixing zone size-LL is computed from mean concentration profile ⟨C⟩H\langle C\rangle_H (Andrews & Spalding 1990) as L=6∫−∞+∞⟨C⟩H(x3,t)(1−⟨C⟩H(x3,t)) dx3L=6\int_{-\infty}^{+\infty}\langle C\rangle_H(x_3,t)(1-\langle C\rangle_H(x_3,t))\,\mathrm{d}x_3. The stratification (or Brunt Väisälä) frequency is defined as

N=−2Ag0∂⟨C⟩H∂x3,(7)N=\sqrt{-2\mathcal{A}g_0\frac{\partial\langle C\rangle_H}{\partial x_3}}, \tag*{(7)}

where the vertical gradient of mean concentration approximated as ∂⟨C⟩H/∂x3=−1/L\partial\langle C\rangle_H/\partial x_3=-1/L.

Theoretical model

Singh & Pal [17] presents the details of the inviscid linear stability analysis of the Faraday instability in a rotating and vertically oscillating two-layer miscible fluid system. Here, we briefly discuss the origin of the derivation of the linearized governing equations (Mathieu equations) along with the assumptions. The equations for the fluctuating velocity u(x,t)\boldsymbol{u}(\boldsymbol{x},t) and the concentration c(x,t)c(\boldsymbol{x},t) fields are obtained by substituting the Reynolds decomposition 2.2a into the governing equations 2.1a, 2.1b and 2.1c, with mean velocity field ⟨U⟩H=0\langle\boldsymbol{U}\rangle_H=0 and ⟨u⟩H=⟨c⟩H=⟨p⟩H=0\langle\boldsymbol{u}\rangle_H=\langle c\rangle_H=\langle p\rangle_H=0. The resulting equations are given as

∂ui∂xi=0,(8)\frac{\partial u_i}{\partial x_i}=0, \tag*{(8)}
∂ui∂t+uj∂ui∂xj=−∂p∂xi+fϵij3uje^3−2Ag(t)cδi3+∂⟨uiu3⟩H∂x3+ν∇2ui,(9)\frac{\partial u_i}{\partial t}+u_j\frac{\partial u_i}{\partial x_j}=-\frac{\partial p}{\partial x_i}+f\epsilon_{ij3}u_j\hat{e}_3-2\mathcal{A}g(t)c\delta_{i3}+\frac{\partial\langle u_i u_3\rangle_H}{\partial x_3}+\nu\nabla^2u_i, \tag*{(9)}
∂c∂t+uj∂c∂xj=−u3∂⟨C⟩H∂x3+∂⟨u3c⟩H∂x3+κ∇2c.(10)\frac{\partial c}{\partial t}+u_j\frac{\partial c}{\partial x_j}=-u_3\frac{\partial\langle C\rangle_H}{\partial x_3}+\frac{\partial\langle u_3c\rangle_H}{\partial x_3}+\kappa\nabla^2c. \tag*{(10)}

We assume that the fluids are inviscid and the fluctuations in velocity and concentration are small. Therefore, the non-linear, the viscous, and the diffusion terms are neglected from the equations 2.4b and 2.4c. The concentration profile is assumed to be piecewise, i.e. constant in the bottom and upper pure (unmixed) fluids and linear in the mixing layer, with a vertical gradient of mean concentration ∂⟨C⟩H/∂x3=0\partial\langle C\rangle_H/\partial x_3=0, in the unmixed region and ∂⟨C⟩H/∂x3=constant\partial\langle C\rangle_H/\partial x_3=\text{constant}, in the mixed region. The detailed derivation of the linearized equations and their solution strategy is presented in Singh & Pal [17]. The final system of linearized equations (for N>fN>f), equivalent to a set of Mathieu equations, is given as

∂2a∂τ2+(f2cos⁡2(θ)ω2+N2sin⁡2(θ)ω2(1+Fcos⁡(τ)))a=0,(11)\frac{\partial^2a}{\partial\tau^2}+\left(\frac{f^2\cos^2(\theta)}{\omega^2}+\frac{N^2\sin^2(\theta)}{\omega^2}(1+F\cos(\tau))\right)a=0, \tag*{(11)}

where, aa represents the amplitude of the concentration field mode, τ=(ωt)\tau=(\omega t) is the non-dimensional time, and θ\theta is the angle between the vertical axis and the wavevector K\mathbf{K} with components (k,l,m)(k,l,m) in the x1x_1, x2x_2 and x3x_3 directions, respectively. The Floquet theory, which provides the solutions to linear differential equations with periodic coefficients, is utilized to solve a set of Mathieu equations 2.5. Details of solving the equation 2.5 using Floquet theory are presented in Appendix A. The solutions of the equation 2.5 can be stable or unstable which governs the stability of the system.

Simulation details

The governing equations 2.1a, 2.1b, and 2.1c are solved in a Cartesian coordinate system on a staggered grid using a finite difference method. The numerical algorithms are presented in Singh & Pal [17]. The numerical solver has been extensively validated and used for studies of rapidly rotating convection-driven dynamos [7 b], rotating convection [12], and several stratified free-shear and wall-bounded turbulent flows [14, 13, 11] (Brucker & Sarkar 2010). A sponge region is used near the top and bottom boundaries to control the spurious reflections from the disturbances propagating out of the domain, where damping functions gradually relax the velocities to their corresponding values at the boundaries. These damping functions are added on the right-hand side of the momentum equation 2.1b as explained in Brucker & Sarkar (2010). The sponge region is sufficiently far away from the mixing region such that it does not affect the dynamics of the mixing of fluids. We consider periodic boundary conditions for all variables in the x1x_1 and x2x_2 (horizontal) directions, whereas at the top and bottom walls (x3x_3), no-slip boundary conditions for velocity vector U\mathbf{U} and Neumann boundary conditions for CC and pressure PP are used. The initial concentration profile [17] (Briard et al. 2019; Briard et al. 2020) in the present simulations is given as

C(x,t=0)=12(1−tanh⁡(2x3δ)),(12)C(\mathbf{x},t=0)=\frac{1}{2}\left(1-\tanh\left(\frac{2x_3}{\delta}\right)\right), \tag*{(12)}

where the parameter δ\delta is used to change the initial mixing zone size-L0L_0. δ=0.035\delta=0.035 and corresponding L0=0.096 mL_0=0.096\ \mathrm{m} is used across all simulations. Broadband fluctuations are imposed on the initial concentration profile 2.6 as reported in Singh & Pal [17]. The domain size in the horizontal directions is lx1=lx2=2πl_{x_1} = l_{x_2} = 2\pi m, and in the vertical direction is lx3=2H=3.5πl_{x_3} = 2H = 3.5\pi m which includes sponge region of thickness Hs=0.64πH_s = 0.64\pi m at top and bottom boundaries. We use uniform grids Nx1=Nx2=512N_{x_1} = N_{x_2} = 512 in the horizontal directions, while in the vertical direction x3x_3, non-uniform grids Nx3=512N_{x_3} = 512 are used with clustering at the vertical centre region of thickness 1.53π1.53\pi m with Δx3min⁡=Δx1=Δx2\Delta x_{3\min} = \Delta x_1 = \Delta x_2. Table 1 shows the key parameters for all simulation cases. We select three cases at each FF depending on the value of f/ωf/\omega, where sub-harmonic instability saturates for f/ω=0f/\omega= 0, and 0.480.48 ((f/ω)2<0.25(f/\omega)^2 < 0.25) and never saturates for f/ω=0.59f/\omega= 0.59 ((f/ω)2>0.25(f/\omega)^2 > 0.25) [17]. We refer each case with a unique name for example F075f/ω48, which indicates that F=0.75F = 0.75 and f/ω=0.48f/\omega= 0.48.

CaseFFω\omega (rad s−1^{-1})ff (s−1^{-1})f/ωf/\omega(f/ω)2(f/\omega)^2ηCU\eta_{CU}
F075f/ω\omega00.750.67000∼0.495\sim0.495
F075f/ω\omega480.750.670.3220.4810.23∼0.363\sim0.363
F075f/ω\omega590.750.670.3960.5910.35never saturate
F1f/ω\omega01.00.7000∼0.421\sim0.421
F1f/ω\omega481.00.70.3380.4820.23∼0.372\sim0.372
F1f/ω\omega591.00.70.4140.5910.35never saturate
F2f/ω\omega02.00.8000∼0.392\sim0.392
F2f/ω\omega482.00.80.3840.480.23∼0.366\sim0.366
F2f/ω\omega592.00.80.4730.5920.35never saturate
F3f/ω\omega03.00.9000∼0.345\sim0.345
F3f/ω\omega483.00.90.4320.480.23∼0.333\sim0.333
F3f/ω\omega593.00.90.5320.5920.35never saturate

Table 1. Simulation parameters: ηCU\eta_{CU} is the cumulative mixing efficiency. For all cases we use Atwood number A=0.01\mathcal{A}=0.01, kinematic viscosity ν=1×10−4 m2 s−1\nu=1 \times10^{-4}\ \mathrm{m^2\,s^{-1}}, diffusion coefficient κ=1×10−4 m2 s−1\kappa=1 \times10^{-4}\ \mathrm{m^2\,s^{-1}}, and g0=10 m s−2g_0=10\ \mathrm{m\,s^{-2}} [17] (Briard et al. 2019).

Results

Linear stability analysis

Figure 1a shows the stability diagram at forcing amplitude F=1F = 1 in the (N/ω)2−(f/ω)2(N/\omega)^2 - (f/\omega)^2 parameter plane, obtained by solving the linearized governing equations (Mathieu equations) (11). The stable (white) regions are in between the unstable red (sub-harmonic) and cyan (harmonic) colored tongues. From the Floquet analysis (see Appendix A, Jordan & Smith [6] and Singh & Pal [17]), the solutions inside the first sub-harmonic tongue are periodic with period T=4πT = 4\pi. Therefore, the solutions grow and oscillate with frequency ω/2\omega/2 because cos⁡(τ)\cos(\tau), where τ=ωt\tau= \omega t, in the equation (11) is periodic with T=2πT = 2\pi and frequency is 2π/T=12\pi/T = 1. However, the solutions grow in the second and third sub-harmonic tongues with frequencies of 3ω/23\omega/2 and 5ω/25\omega/2, respectively. The solutions inside the first harmonic tongue are 2π2\pi periodic and grow with frequency ω\omega, whereas, in the second and third harmonic tongues, the solutions grow with frequencies of 2ω2\omega and 3ω3\omega, respectively. The solutions that lie outside the harmonic and sub-harmonic instability tongues are stable. The contour curves for the growth rates (σ\sigma), range from σ=0.02\sigma= 0.02 to σ=0.12\sigma= 0.12 with a step size of 0.020.02, of the unstable θ\theta-modes inside the first leftmost sub-harmonic tongue are illustrated in figure 1b. The black curve corresponds to the neutral stability curve for σ=0\sigma= 0, which separates the regions of stable and unstable solutions. The inclined blue line (solid) segment in figure 1a demonstrates the possible frequencies (∈[(f/ω)2,(N0/ω)2]\in[(f/\omega)^2, (N_0/\omega)^2]) that can be excited corresponding to an angle θ∈[0,π/2]\theta\in[0,\pi/2], for a given (f/ω)2=0.23(f/\omega)^2 = 0.23 and initial stratification condition (N0/ω)2(N_0/\omega)^2 in the first harmonic tongue. This segment ranges from (f/ω)2(f/\omega)^2 (left end (×) at θ=0\theta= 0) to (N0/ω)2(N_0/\omega)^2 (right end (×) at θ=π/2\theta= \pi/2). The unstable θ\theta-modes inside the sub-harmonic or harmonic tongues trigger the Faraday instability. This results in the growth of mixing zone size-LL and decrease of (N/ω)2(N/\omega)^2, as indicated by the arrow directing to the left since (N/ω)2=2Ag0/(Lω2)∝1/L(N/\omega)^2 = 2\mathcal{A}g_0/(L\omega^2) \propto1/L from equation 2.3. Therefore, the condition for the occurrence of instabilities [17] (Gréa & Adou 2018) is defined based on the fact that if at least one θ\theta-mode falls in any unstable tongues solution will grow, resulting in instability. This condition is given as

(Nω)2>G(f,F),(13)\left(\frac{N}{\omega}\right)^2 > \mathcal{G}(f,F), \tag*{(13)}

where G(f,F)\mathcal{G}(f,F) denotes the extreme left boundary of the first leftmost sub-harmonic tongue (see figure 1a). With time evolution, LL grows due to unstable θ\theta-modes and the right end ((N/ω)2(N/\omega)^2) of the blue line (solid) segment (figure 1a) enters the first sub-harmonic tongue as indicated by the blue × at (N/ω)2=0.4(N/\omega)^2 = 0.4 in figure 1b. At this instant, many frequencies (∈[(f/ω)2,(N/ω)2]\in[(f/\omega)^2, (N/\omega)^2]) are excited, corresponding to an angle θ∈[0,π/2]\theta\in[0,\pi/2] and growth rates σ\sigma, as depicted by the inclined dashed blue segment in figure 1b. The sub-harmonic instabilities are triggered in this regime resulting in the intense growth of LL (Singh & Pal 2022). The process repeats, and (N/ω)2(N/\omega)^2 moves to 0.3, as indicated by the dash-dotted blue segment. The instability finally saturates when (N/ω)2(N/\omega)^2 crosses the extreme left neutral stability curve (black) i.e., (N/ω)2=G(f,F)(N/\omega)^2 = \mathcal{G}(f,F). The instability condition 3.1 is no longer fulfilled, and LL saturates. The horizontal green and inclined orange segments (both dash and dash-dotted) represent the cases without rotation (f/ω)2=0(f/\omega)^2 = 0 and with rotation (f/ω)2=0.35(f/\omega)^2 = 0.35, respectively. We note here that the instability saturates for (f/ω)2<0.25(f/\omega)^2 < 0.25.

An interesting feature for (f/ω)2≥0.25(f/\omega)^2 \ge0.25 is observed when (N/ω)2(N/\omega)^2 moves into the leftmost stable regime as indicated by the dotted orange segment (inclined). This segment must always pass through the first sub-harmonic tongue. The instability condition 3.1 is always satisfied, and LL will continue to grow. Therefore, the instability will never saturate.

Figure 1c depicts the possible number of unstable θ\theta-modes permitted to grow at various growth rates σ\sigma, when (N/ω)2=0.4(N/\omega)^2 = 0.4 for each (f/ω)2=0(f/\omega)^2 = 0, 0.230.23, and 0.350.35 cases. Here, the green, blue, and orange bars correspond to the dashed green, blue, and orange segments are shown in Figure 1b. For a given (f/ω)2(f/\omega)^2 and (N/ω)2(N/\omega)^2, we assume a step size of 0.1 between the θ\theta angles while calculating the number of θ\theta-modes between different growth rates with 0.02 size intervals, except for the last interval, which spans from 0.12 to 0.1268 and contains the maximum growth rate. In the non-rotating case (f/ω)2=0(f/\omega)^2 = 0, the growth rate of the fastest growing θ\theta-modes is 0.12≤σ≤0.12680.12 \le\sigma\le0.1268 (see Figure 1c). However, the growth rate of the fastest growing modes in the rotating case (f/ω)2=0.23(f/\omega)^2 = 0.23 (0.08≤σ<0.10.08 \le\sigma< 0.1) is less compared to the non-rotating case. Therefore, the mixing zone size-LL will increase (or (N/ω)2(N/\omega)^2 decrease) gradually for (f/ω)2=0.23(f/\omega)^2 = 0.23, than the rapid increase of LL (or decrease of (N/ω)2(N/\omega)^2) for the non-rotating case. This results in a longer sub-harmonic instability phase for rotating case (f/ω)2=0.23(f/\omega)^2 = 0.23 as compared to the non-rotating case. The sub-harmonic instability phase is the time interval between the onset and the saturation of the sub-harmonic instability. At a higher rotation rate (f/ω)2=0.35(f/\omega)^2 = 0.35, fewer fastest growing θ\theta-modes have a growth rate 0.8≤σ<0.10.8 \le\sigma< 0.1 as compared to (f/ω)2=0.23(f/\omega)^2 = 0.23 (see a comparison among blue and orange bars in Figure 1c). This demonstrates that LL increases much more gradually for (f/ω)2=0.35(f/\omega)^2 = 0.35 than (f/ω)2=0.23(f/\omega)^2 = 0.23. Figure 1d illustrates that at a later instant, when (N/ω)2=0.3(N/\omega)^2 = 0.3 and (f/ω)2=0(f/\omega)^2 = 0, only a small number of θ\theta-modes are permitted to grow at smaller growth rates σ<0.08\sigma< 0.08, whereas a large number of unstable modes grows at larger growth rates σ>0.08\sigma> 0.08. The scenario is different for (f/ω)2=0.23(f/\omega)^2 = 0.23, where a significant number of unstable modes are permitted to grow at all growth rates. This results in the triggering of more sub-harmonic instabilities and suggests an increase in turbulent kinetic energy (t.k.e.t.k.e.) for cases with 0<(f/ω)2<0.250 < (f/\omega)^2 < 0.25 as compared to the non-rotating case since sub-harmonic instabilities are responsible for the turbulent mixing [17]. Although, more unstable modes are permitted to grow at growth rates upto σ<0.12\sigma< 0.12 for (f/ω)2=0.35(f/\omega)^2 = 0.35 compared to (f/ω)2=0(f/\omega)^2 = 0, a significantly less number of fastest growing modes grow at 0.12≤σ≤0.12680.12 \le\sigma\le0.1268. This suggests that higher rotation rates, (f/ω)2≥0.25(f/\omega)^2 \ge0.25, have weaker turbulence.

Image not extracted

Figure 1. (a) Stability diagram for the solutions to the Mathieu equation (11) [17] at forcing amplitude F=1F = 1. The stable (white) regions are in between the unstable red (sub-harmonic) and cyan (harmonic) colored tongues. The inclined blue line (solid) segment represents the frequencies excited, corresponding to an angle θ∈[0,π/2]\theta\in[0, \pi/2], for a given (f/ω)2=0.23(f/\omega)^2 = 0.23 (left end (×) at θ=0\theta= 0), and initial condition (N0/ω)2(N_0/\omega)^2 (right end (×) at θ=π/2\theta= \pi/2) in the first harmonic tongue. Arrow indicates the evolution of (N/ω)2(N/\omega)^2 as the mixing zone size-LL increases. Panel (b) is an enlargement of the first leftmost sub-harmonic tongue in (a). The black curve denotes the neutral stability curve for the growth rate σ=0\sigma= 0. The color curves respectively denote the contours of different growth rates (σ\sigma) at intervals of 0.02. The dashed and dash-dotted green, blue and orange segments demonstrate the non-rotating case (f/ω)2=0(f/\omega)^2 = 0 and rotating cases (f/ω)2=0.23(f/\omega)^2 = 0.23 and 0.35, respectively, when (N/ω)2(N/\omega)^2 reaches to 0.4 and 0.3 during the growth of LL. The possible number of unstable θ\theta-modes permitted to grow at various growth rates (σ\sigma) for each (f/ω)2=0(f/\omega)^2 = 0, 0.23, and 0.35 cases are shown in (c) when (N/ω)2=0.4(N/\omega)^2 = 0.4, and (d) when (N/ω)2=0.3(N/\omega)^2 = 0.3.

Figure 2a illustrates the stability diagram at forcing amplitude F=3F = 3. The contour curves for the growth rates, from σ=0\sigma= 0 to σ=0.28\sigma= 0.28 with a step size of 0.04, in the first leftmost sub-harmonic tongue, are shown in Figure 2b. We can observe that with an increase in FF, the first sub-harmonic and harmonic tongues become wider, and the stable region between these tongues shrinks resulting in the early onset of the sub-harmonic instability and turbulent mixing similar to the non-rotating case [17]. Additionally, for all (f/ω)2(f/\omega)^2 cases, the maximum growth rate (σmax\sigma_{max}) of the fastest growing modes is 2.7 times greater at F=3F = 3 than at lower forcing amplitude (F=1F = 1) i.e., σmax=0.338\sigma_{max} = 0.338 at F=3F = 3, whereas σmax=0.1268\sigma_{max} = 0.1268 at F=1F = 1. As a result, the sub-harmonic instability is triggered earlier at F=3F = 3 than for F=1F = 1. This suggests that the sub-harmonic instability phase at F=3F = 3 is shorter than the sub-harmonic instability phase at F=1F = 1 for (f/ω)2<0.25(f/\omega)^2 < 0.25 cases. Similar to the cases at F=1F = 1, the instability saturates for (f/ω)2<0.25(f/\omega)^2 < 0.25, whereas the instability never saturates for (f/ω)2≥0.25(f/\omega)^2 \ge0.25 and LL continues to grow at F=3F = 3, owing to the continuous triggering of the unstable θ\theta-modes as demonstrated by the dotted orange segment in Figure 2b. These predictions from the linear stability analysis will further support our numerical simulations as presented in the next section.

Figure 2. (a) Stability diagram for the solutions to the Mathieu equation (11) [17] at F=3F = 3. Panel (b) is the enlargement of the first leftmost unstable sub-harmonic (red) tongue of panel (a). The unstable sub-harmonic region is bounded by the black neutral stability curve for the growth rate σ=0\sigma= 0 and the color curves represent the contours of different growth rates (σ\sigma) at intervals of 0.04. The dotted orange segment demonstrate the rotating case (f/ω)2=0.35(f/\omega)^2 = 0.35. The arrow indicates the evolution of (N/ω)2(N/\omega)^2 as the mixing zone size-LL increases.

Numerical simulations

The evolution of the vertically (x3x_3) integrated turbulent kinetic energy (t.k.e.t.k.e.) for F=1F = 1 is illustrated in Figure 3a. The t.k.e.t.k.e. is defined as

Figure 3. Evolution of vertically (x3x_3) integrated t.k.e.t.k.e. for both without (f/ω=0f/\omega= 0) and with (f/ω=0.48,0.59f/\omega= 0.48, 0.59) rotation cases at (a) F=1F = 1, and (b) F=3F = 3. Inset shows the evolution of mixing zone size-LL with shaded regions indicating the sub-harmonic instability phase. The vertical red dashed line in the inset of panel (a) denotes the onset of sub-harmonic instability for F1f/ωf/\omega59.

t.k.e.=⟨u12⟩H+⟨u22⟩H+⟨u32⟩H2,u=U−⟨U⟩H.(14)t.k.e. = \frac{\langle u_1^2\rangle_H + \langle u_2^2\rangle_H + \langle u_3^2\rangle_H}{2}, \qquad\boldsymbol{u} = \boldsymbol{U} - \langle\boldsymbol{U}\rangle_H . \tag*{(14)}

Here, ⟨U⟩H(x3,t)\langle\boldsymbol{U}\rangle_H(x_3,t) is the horizontally averaged mean velocity field and u(x1,x2,x3,t)\boldsymbol{u}(x_1,x_2,x_3,t) is the fluctuating velocity field. The t.k.e.t.k.e. rapidly grows for F1f/ω\omega0 and F1f/ω\omega48 during the onset of the sub-harmonic instability. Eventually, the t.k.e.t.k.e. decays owing to the instability saturation. The t.k.e.t.k.e. for F1f/ω\omega48 is higher than F1f/ω\omega0 due to the excitement of more unstable θ\theta-modes in the sub-harmonic region that are permitted to grow with growth rates (σ\sigma) ranging from 0 to 0.1268, as observed from the stability analysis. The t.k.e.t.k.e. decreases significantly for f/ω=0.59f/\omega= 0.59 (case F1f/ω\omega59) due to the effect of strong rotation The excitement of a small number of unstable θ\theta-modes at higher growth rates, as predicted from the linear stability analysis, is also a possible reason for the low t.k.e.t.k.e. for F1f/ω\omega59. Although small, the t.k.e.t.k.e. for F1f/ω\omega59 sustains owing to the continuous triggering of the sub-harmonic instability, resulting in continuous turbulent mixing. A similar evolution of t.k.e.t.k.e. is observed for F=0.75F = 0.75. At a higher forcing amplitude of F=3F = 3, turbulence appears at the beginning, and therefore, t.k.e.t.k.e. is similar for all f/ωf/\omega cases (Figure 3b), indicating that the instability delaying effect of rotation is mitigated by the strong vertical forcing. Notice that the maximum value of t.k.e.t.k.e. for F1f/ω\omega48 is higher than F3f/ω\omega48. This observation is also true for the non-rotating cases at the respective forcing amplitudes. The longer sub-harmonic instability phase for F1f/ω\omega48 is the reason for a higher value of t.k.e.t.k.e. at F1f/ω\omega48 than at F3f/ω\omega48. The sub-harmonic instability phase is demonstrated using the shaded regions in the evolution of the mixing zone size LL in the inset of Figure 3a and Figure 3b.

A longer sub-harmonic instability phase for F=0.75,1F = 0.75, 1 signifies a delay in the saturation of the instabilities triggered during the low amplitude oscillations. Therefore, the instabilities get sufficient time to evolve, eventually resulting in turbulence. In contrast to F=0.75,1F = 0.75, 1, at F=2,3F = 2, 3, the sub-harmonic instability phase is short. The instabilities saturate quickly and therefore do not get sufficient time to evolve and intensify turbulence. This observation is consistent with our linear stability analysis that predicts the maximum growth rate of the fastest growing modes is 2.7 times greater at F=3F = 3 than F=1F = 1, resulting in the shorter sub-harmonic instability phase at F=3F = 3. Interestingly, both F2f/ωf/\omega59 (figure not shown) and F3f/ωf/\omega59 manifest a recovery in t.k.e.t.k.e. owing to the continuous triggering of the sub-harmonic instabilities.

The transition to turbulence phenomenon for the cases without and with rotation is analyzed using the contour plots of the concentration fields in a vertical plane. Figure 4 depicts the evolution of the interface of the concentration field for F1f/ωf/\omega0. The initially perturbed interface, subjected to vertical and periodic accelerations, deforms into a wavy form owing to harmonic and sub-harmonic resonance as shown in Figure 4a at ωt=33.6\omega t = 33.6. This diffused wavy interface is amplified by the sub-harmonic instability, leading to the formation of mushroom-shaped waves as shown in Figure 4b at ωt=49\omega t = 49. At this instance, some waves with shorter wavelengths become disorganized and interact with each other, resulting in small velocity fluctuations. In the next oscillation, the amplitude of a longer wavelength wave amplifies, leading to the birth of a well-defined mushroom-shaped wave that bursts at the nodes as depicted in Figure 4c at ωt=56\omega t = 56. This process is known as wave-breaking of the Faraday waves (Cavelier et al. 2022) and is responsible for the onset of irreversible mixing. In the successive oscillations, the entire interface breaks down into small-scale structures leading to a full transition to turbulence as manifested in Figure 4d and Figure 4e. A similar dynamics is observed for F=0.75F=0.75 and f/ω=0f/\omega=0 (see Movie 1 and Movie 4).

Figure 4. Contours of concentration field in vertical x1−x3x_1-x_3 center plane (x2=πx_2=\pi) for case F1f/ω0F1f/\omega0 at time instants (a) ωt=33.6\omega t=33.6, (b) ωt=49\omega t=49, (c) ωt=56\omega t=56, (d) ωt=61.6\omega t=61.6, and (e) ωt=67.2\omega t=67.2. Pure fluids of C=1C=1 (denser) and C=0C=0 (lighter) are made transparent.

The combined effect of vertical oscillation and rotation in F1f/ω48F1f/\omega48 results in roll-ups at multiple locations on the diffuse interface as shown in Figure 5a at ωt=175.28\omega t=175.28. In subsequent oscillations, roll-ups will continue to form with a large size but in the opposite direction to that of the previous one. These roll-ups will finally break into turbulence (see Figure 5a). Notice that the entire interface layer breaks into small structures in one oscillation resulting in a larger thickness of the turbulent mixing region as compared to F1f/ω0F1f/\omega0 where the mushroom-shaped wave first breaks at the nodes and then turbulence spreads to the entire interface layer in the subsequent oscillations. Movie 2 included as the supplementary data demonstrates this event. The formation of the roll-ups and their eventual breaking to turbulence is responsible for larger t.k.e.t.k.e. in F1f/ω48F1f/\omega48 as compared to F1f/ω0F1f/\omega0 (Figure 3a). This phenomenon is consistent with our linear stability analysis as follows. For F1f/ω48F1f/\omega48 a significant number of unstable modes are permitted to grow at all growth rates. This has been demonstrated in Figure 1c and Figure 1d when (N/ω)2=0.4(N/\omega)^2=0.4 and 0.30.3 respectively. The ωt\omega t corresponding to (N/ω)2=0.4(N/\omega)^2=0.4 and 0.30.3 is in the range 175−205175-205. Figure 5a shows the formation of the roll-ups at ωt=193.2\omega t=193.2, and therefore it can be deduced that the unstable modes promote the formation of roll-ups that break to enhance t.k.e.t.k.e. for F1f/ω48F1f/\omega48. We observe similar dynamics for F075f/ω48F075f/\omega48 (see Movie 5). This phenomenon results in a gradual increase of irreversible mixing. When the Coriolis frequency increase to f/ω=0.59f/\omega=0.59 (case F1f/ω59F1f/\omega59), we found that the effect of rotation is strong enough to suppress the formation of large amplitude waves and roll-ups. Therefore, the turbulent mixing is only caused by the finger-shaped instability as depicted in Figure 5b (see supplementary Movie 3 and Movie 6).

Contours of concentration field in vertical $x_2-x_3$ center plane for rotation cases

Figure 5. Contours of concentration field in vertical x2−x3x_2-x_3 center plane (x1=πx_1=\pi) for rotation cases (a) F1f/ω\omega48, (b) F1f/ω\omega59, and (c) F3f/ω\omega48 at different time instants during the instability growth stage.

At higher forcing amplitude F=3F=3 for f/ω=0.48f/\omega=0.48, we observe the development of random finger-shaped structures on the interface that break in the first oscillation as shown in the Figure 5c. Therefore, the turbulent mixing starts from the beginning of the periodic forcing. Additionally, the intensity of turbulence is low because of the shorter sub-harmonic instability phase at higher FF. As the instability delaying effect of rotation is subdued at higher forcing amplitudes [17], the turbulent mixing occurs by this mechanism for all f/ωf/\omega at F=2,3F=2,3 (figures not shown). We have included the animation of the concentration field for F3f/ω48F3f/\omega48 as supplementary data (Movie 7).

Since there is no mean velocity field involved, we start the analysis of the irreversible mixing using the t.k.e.t.k.e. evolution equation as follows:

d t.k.e.dt=P−S−B−ϵ−∂Tj∂xj.(15)\frac{d\,t.k.e.}{dt}=\mathcal{P}-S-B-\epsilon-\frac{\partial\mathcal{T}_j}{\partial x_j}. \tag*{(15)}

The turbulent production term P=−⟨uiuj⟩H∂⟨Ui⟩H∂xj\mathcal{P}=-\langle u_i u_j\rangle_H\frac{\partial\langle U_i\rangle_H}{\partial x_j} and the transport term ∂Tj/∂xj\partial\mathcal{T}_j/\partial x_j, where Tj=⟨puj⟩H−2ν⟨sijui⟩H+12⟨uiuiuj⟩H\mathcal{T}_j=\langle p u_j\rangle_H-2\nu\langle s_{ij}u_i\rangle_H+\frac{1}{2}\langle u_i u_i u_j\rangle_H are negligible. The energy input from the periodic forcing is S=2Ag0Fcos⁡(ωt)⟨u3c⟩HS=2\mathcal{A}g_0F\cos(\omega t)\langle u_3c\rangle_H. The buoyancy flux B=2Ag0⟨u3c⟩HB=2\mathcal{A}g_0\langle u_3c\rangle_H accounts for the reversible rate of exchange between t.k.e.t.k.e. and the potential energy. The viscous dissipation ϵ=2ν⟨sijsij⟩H\epsilon=2\nu\langle s_{ij}s_{ij}\rangle_H, where sij=(∂ui/∂xj+∂uj/∂xi)/2s_{ij}=(\partial u_i/\partial x_j+\partial u_j/\partial x_i)/2, acts as a sink for t.k.e.t.k.e.. We obtain an excellent closure of equation 3.3 for all the simulations signifying sufficient grid spacing in all directions to resolve all length scales.

The irreversible mixing is characterized by partitioning the total potential energy (PE) Ep=2Ag0∫VC(x)x3 dVE_p=2\mathcal{A}g_0\int_V C(\boldsymbol{x})x_3\,dV into the sum of an available potential energy (APE) EaE_a and background potential energy (BPE) EbE_b [20]. The BPE is the minimum potential energy a flow can have and is unavailable for extraction from the potential energy reservoir for driving macroscopic fluid motion. To estimate BPE of the flow at any instance of time, a volume element dVdV with concentration CC at height x3x_3 is adiabatically relocated to x3∗x_3^* such that the redistributed concentration field is statistically stable (dC(x3∗)/dx3∗≤0dC(x_3^*)/dx_3^*\leq0) everywhere. We compute Eb=2Ag0∫VC(x3∗)x3∗ dVE_b=2\mathcal{A}g_0\int_V C(x_3^*)x_3^*\,dV and subtract it from EpE_p to obtain EaE_a [19] (Chalamalla & Sarkar 2015). EaE_a is available to be converted into kinetic energy. Since our domain is closed by the top and bottom boundaries, the surface fluxes are zero. We derive the simplified equations for the evolution of PE, BPE, and APE following Winters et al. [20],

dEpdt=BV+ϕi,(16)\frac{dE_p}{dt}=B_V+\phi_i, \tag*{(16)}
dEbdt=ϕd,(17)\frac{\mathrm{d}E_b}{\mathrm{d}t}=\phi_d, \tag*{(17)}
dEadt=BV−(ϕd−ϕi).(18)\frac{\mathrm{d}E_a}{\mathrm{d}t}=B_V-(\phi_d-\phi_i). \tag*{(18)}

Here BVB_V, ϕi\phi_i, and ϕd\phi_d are integrated over the volume V=lx1lx2(2H−2Hs)V=l_{x_1}l_{x_2}(2H-2H_s), and (2H−2Hs)(2H-2H_s) is the vertical height of the domain excluding top and bottom sponge layer thickness. The BV=∫Blx1lx2 dx3B_V=\int B l_{x_1}l_{x_2}\,\mathrm{d}x_3 is volume integrated buoyancy flux and ϕi=−2Ag0κA(⟨Ctop⟩H−⟨Cbot⟩H)\phi_i=-2A g_0\kappa A\left(\langle C_{top}\rangle_H-\langle C_{bot}\rangle_H\right), where A=lx1lx2A=l_{x_1}l_{x_2}, the rate at which the potential energy of a statically stable concentration (CC) distribution increases in the absence of macroscopic fluid motion through the irreversible conversion of internal to potential energy. The ϕi\phi_i is a strictly non-negative quantity and signifies the increase in BPE at a constant rate ϕi\phi_i owing to the laminar diffusion. ϕd=2Ag0κ∫V−dx3∗dC∣∇C∣2dV\phi_d=2A g_0\kappa\int_V\frac{-\mathrm{d}x_3^*}{\mathrm{d}C}\lvert\nabla C\rvert^2\mathrm{d}V represents the rate of change of BPE due to irreversible diapycnal flux and is a measure of the irreversible mixing. Notice that as ϕd≥0\phi_d\geq0, EbE_b always increases as a result of diapycnal mixing. ϕd\phi_d is associated with the increase in BPE due to laminar diffusion and energy coming directly from the t.k.e.t.k.e. reservoir via BVB_V which is related to the turbulent mixing. We wish to investigate the energetics associated with the turbulent mixing caused by the sub-harmonic instability. Therefore, we define the irreversible mixing rate M\mathcal{M} (=ϕd−ϕi=\phi_d-\phi_i) to isolate the increase in BPE due to purely turbulent mixing from that of laminar diffusion [15]. Since we have periodic forcing in the domain, we choose to calculate the cumulative mixing efficiency (Peltier & Caulfield 2003; Briard et al. 2019) as follows:

ηcu(t)=∫0tM dt∫0t(M+ϵV) dt.(19)\eta_{cu}(t)=\frac{\int_0^t\mathcal{M}\,\mathrm{d}t}{\int_0^t(\mathcal{M}+\epsilon_V)\,\mathrm{d}t}. \tag*{(19)}

Here ϵV=∫ϵlx1lx2 dx3\epsilon_V=\int\epsilon l_{x_1}l_{x_2}\,\mathrm{d}x_3 is the volume integrated viscous dissipation. The cumulative mixing efficiency ηcu\eta_{cu} measures the irreversible loss of t.k.e.t.k.e. that expends in irreversible turbulent mixing in contrast to the total t.k.e.t.k.e. loss to the irreversible turbulent mixing and viscous dissipation.

Figure 6 illustrates the energy pathways. Arrows denote the energy exchanges between different energy reservoirs. SVS_V depicts the input rate of external energy to the t.k.e. reservoir. In the presence of fluid motion due to turbulent mixing, t.k.e. enters the potential energy reservoir via reversible buoyancy flux BVB_V, increasing EaE_a. Some part of EaE_a can go to EbE_b via purely turbulent mixing as quantified by M\mathcal{M}, resulting in irreversible mixing. The remaining portion of EaE_a converts back to t.k.e. by BVB_V. The energy loss from the t.k.e. reservoir to the internal energy reservoir occurs through viscous dissipation ϵV\epsilon_V. The laminar diffusion contributes energy to EbE_b directly via ϕi\phi_i.

Schematic of energy pathways

Figure 6. Schematic of energy pathways: SVS_V is the volume integrated SS denoting the external input of energy from periodic forcing, BVB_V is reversible buoyancy flux, M\mathcal{M} is irreversible mixing rate due to purely turbulent mixing, ϕd\phi_d is irreversible diapycnal flux, ϕi\phi_i is conversion of internal to potential energy due to laminar diffusion, and ϵV\epsilon_V is kinetic energy dissipation.

Figure 7 illustrates the temporal evolution of the reversible buoyancy flux BVB_V and the irreversible mixing rate M\mathcal{M}. We also include the evolution of Ep′E'_p, Eb′E'_b, and Ea′E'_a in the inset of each figure. Ep′E'_p and Eb′E'_b demonstrate the total change in PE and BPE due to irreversible turbulent mixing and are calculated relative to their initial values. Ea′(=Ep′−Eb′)E'_a(=E'_p-E'_b) denotes the corresponding change in APE. The cumulative energy changes in PE and BPE associated with the constant rate of laminar diffusion are not included in Ep′E'_p and Eb′E'_b. Therefore, Ep′(t)=Ep(t)−Ep(0)−ϕitE'_p(t)=E_p(t)-E_p(0)-\phi_i t and Eb′(t)=Eb(t)−Eb(0)−ϕitE'_b(t)=E_b(t)-E_b(0)-\phi_i t, where tt is the time from the start of the simulation. The t.k.e. for F=1,f/ω=0F=1,f/\omega=0 increases due to the onset of sub-harmonic instabilities and the formation and breaking of mushroom-shaped waves. This increased t.k.e. results in an increase in BVB_V at ωt≃46−55\omega t \simeq46-55, and therefore, the potential energy of the flow also increases as demonstrated in Figure 7a. The PE is stored as APE, which almost returns to t.k.e.t.k.e. via BVB_V because the transfer of APE to BPE remains zero as quantified by M≈0\mathcal{M} \approx0. In the subsequent oscillation, this t.k.e.t.k.e. contributes to the external energy input, resulting in the increased t.k.e.t.k.e. and wave-breaking. The M\mathcal{M} also starts increasing after ωt≃55\omega t \simeq55. After wave-breaking, the entire interface layer breaks, resulting in a rapid increase of M\mathcal{M} signifying rapid conversion of APE into BPE, and the APE decreases. When the instabilities saturate at ωt≃70\omega t \simeq70, t.k.e.t.k.e., and BVB_V decrease, leading to a decrease in APE and PE saturates. The BPE also saturates, as demonstrated by a decreasing M\mathcal{M}. Similar to F1f/ω0\omega0, we observe that BVB_V for F1f/ω48\omega48 increases at ωt≃180\omega t \simeq180 owing to an increase in t.k.e.t.k.e. (see Figure 7b). At the same instance, BVB_V, PE, and APE also increase. We find a gradual increase of M\mathcal{M} and BPE for F1f/ω48\omega48 compared to a rapid increase for F1f/ω0\omega0. The reason is the continuous formation and breaking of the roll-ups at the diffuse interface during the sub-harmonic instability phase for F1f/ω48\omega48. With the saturation of the instability, BVB_V and APE start decreasing, resulting in the increase and eventual saturation of BPE as illustrated by the gradual decrease of M\mathcal{M} to zero in Figure 7b. This gradual decrease in M\mathcal{M} denotes that irreversible mixing is sustained for an extended period than F1f/ω0\omega0, owing to the long sub-harmonic instability phase. For F1f/ω59\omega59, there is a significant delay in the onset of sub-harmonic instability due to the presence of the stable region where no harmonic or sub-harmonic modes are excited ([17]). This stable region is apparent in the evolution of mixing zone size-LL for F1f/ω59\omega59 as shown in the inset of Figure 3a. Therefore, BVB_V and APE ∼0\sim0 till ωt≃280\omega t \simeq280. During this period, the initial concentration profile diffuses while maintaining its static stability (Winters et al. [20]). The flow remains in its reference state of minimum potential energy till the onset of turbulent mixing, signifying Ep=EbE_p = E_b from the definition of EbE_b. No APE is created and ϕd=ϕi\phi_d = \phi_i (from equations 3.4a, 3.4b and 3.4c) signifying the increase in BPE via ϕi\phi_i. Therefore, the potential energy increases at the expense of the internal energy of the fluid. We find, ϕd(0.001)≈ϕi(0.0008)\phi_d(0.001) \approx\phi_i(0.0008) till ωt≃280\omega t \simeq280. A similar phenomenon is also observed for F1f/ω48\omega48, and for all f/ωf/\omega at F=0.75F = 0.75. For higher rotation rate the turbulence is weaker (see Movie 3), resulting in smaller t.k.e.t.k.e. and BVB_V (Figure 7c) compared to the previous cases. BVB_V increases after ωt≃280\omega t \simeq280, increasing APE. The major portion of the APE coming from t.k.e.t.k.e. via BVB_V is transferred to BPE via M\mathcal{M}. As a result, only a minor portion of the APE is available for reversible exchange back to t.k.e.t.k.e., as shown by small APE in the inset of Figure 7c. Therefore, in the subsequent oscillation, only a small amount of kinetic energy is available in the t.k.e.t.k.e. reservoir, which contributes to the external energy input, resulting in weaker turbulence. The instability never saturates for F1f/ω59\omega59 and triggers continuously, resulting in the sustenance of t.k.e.t.k.e. and BVB_V. The PE also continues to increase due to BVB_V, and APE continues to remain small since much of the APE goes to BPE. M\mathcal{M} oscillates and remains greater than zero indicating a continuous increase in BPE with time and signifying the sustenance of irreversible mixing for higher rotation rates.

Figure 7. Time evolution of buoyancy flux (BVB_V) and irreversible mixing rate (M\mathcal{M}) at F=1F=1 for (a) f/ω=0f/\omega=0, (b) f/ω=0.48f/\omega=0.48, (c) f/ω=0.59f/\omega=0.59, and at F=3F=3 for (d) f/ω=0.59f/\omega=0.59. The corresponding time evolution of change in PE (Ep′(t)=Ep(t)−Ep(0)−ϕitE'_p(t)=E_p(t)-E_p(0)-\phi_i t), BPE (Eb′(t)=Eb(t)−Eb(0)−ϕitE'_b(t)=E_b(t)-E_b(0)-\phi_i t), and APE (Ea′(t)=Ep′(t)−Eb′(t)E'_a(t)=E'_p(t)-E'_b(t)), due to irreversible turbulent mixing are shown in the inset of each panel.

For higher forcing amplitudes F=2,3F = 2, 3 and f/ω=0,0.48f/\omega= 0, 0.48, the t.k.e.t.k.e. and BVB_V increase from the beginning of the periodic forcing, resulting in the increase of APE (figure not shown). Subsequently, BPE and M\mathcal{M} also increase. Since the instabilities saturate quickly for these cases, BVB_V and APE become approximately zero. The BPE saturates, and the M\mathcal{M} also tends to zero, signifying the ceasing of irreversible mixing. The initial evolution of energetics for F=2,3F = 2, 3 for f/ω=0.59f/\omega= 0.59 (Figure 7d for F3f/ω59\omega59) is similar to the cases discussed above. However, at a later stage, the evolution resembles the case F1f/ω59\omega59 and demonstrates non-zero BVB_V and M\mathcal{M}, signifying a continuous increase in BPE and sustenance of irreversible turbulent mixing.

The time evolution of the cumulative mixing efficiency ηcu\eta_{\mathrm{cu}} is plotted in Figure 8. We report the ηcu\eta_{\mathrm{cu}} values when the instability saturates for f/ω=0f/\omega= 0 and 0.480.48 cases. Since the instability never saturates for f/ω=0.59f/\omega= 0.59, the ηcu\eta_{cu} is given when M\mathcal{M} decreases to its approximately minimum value after reaching its peak. For F075f/ω\omega0 case, ηcu\eta_{cu} decreases rapidly when the irreversible mixing rate M\mathcal{M} and the viscous dissipation ϵV\epsilon_V increases. It attains a minimum value of ∼0.495\sim0.495 after instability saturation. In the presence of rotation for F075f/ω\omega48, the ηcu\eta_{cu} reduces to ∼0.363\sim0.363, which implies that the mixing is not as efficient as for F075f/ω\omega0. For F075f/ω\omega59, the ηcu\eta_{cu} decreases continuously owing to the continuous irreversible mixing. However, the mixing is still more efficient (ηcu≈0.52\eta_{cu} \approx0.52 at ωt≈825\omega t \approx825) than F075f/ω\omega0 and F075f/ω\omega48. For F1f/ω\omega0, ηcu\eta_{cu} is ∼0.421\sim0.421 similar to that obtained by Briard et al. (2019). At F=1F = 1 and f/ω=0.48f/\omega= 0.48 ηCu∼0.372\eta_{Cu} \sim0.372 whereas for f/ω=0.59f/\omega= 0.59 the cumulative mixing efficiency is 0.406 at ωt=460\omega t = 460, but keeps decreasing with time. We find that the mixing becomes less efficient with an increase in FF for all f/ωf/\omega values, and the mixing efficiency continues to decrease for the f/ω=0.59f/\omega= 0.59 cases. In the present work, the primary mechanism for mixing vertically oscillating stably stratified fluids is the Faraday instability. In density-stratified shear flows the Kelvin-Helmholtz instability is responsible for mixing. Despite this difference in the mixing mechanisms, the cumulative mixing efficiencies obtained in our simulations are surprisingly similar to those of the stratified shear turbulent flows ([5]).

Figure 8. Time evolution of cumulative mixing efficiency ηcu\eta_{cu} for each without rotation (f/ω=0f/\omega= 0) and with rotation (f/ω=0.48f/\omega= 0.48, 0.59) cases at forcing amplitudes (a) F=0.75F = 0.75, (b) F=1F = 1, (c) F=2F = 2, and (d) F=3F = 3.

Conclusions

We perform DNS to investigate the effect of rotation on the turbulent mixing driven by the Faraday instability in two miscible fluids subjected to vertical oscillations. At lower forcing amplitudes F=0.75F = 0.75 and 1, the t.k.e.t.k.e. increases with an increase in the Coriolis frequency from f/ω=0f/\omega= 0 to 0.48. This enhancement is attributed to the excitement of more unstable θ\theta-modes in the sub-harmonic region that results in the development and breaking of roll-ups at the interface layer during the sub-harmonic instability phase. The mechanism of turbulence generation for the cases without rotation is different. A well-defined mushroom-shaped wave evolves for f/ω=0f/\omega= 0 cases that disintegrate at the nodes. Subsequently, the turbulence spreads to the entire mixing layer in successive oscillations. In contrast to f/ω=0f/\omega= 0, 0.48, f/ω=0.59f/\omega= 0.59 have a weaker turbulence owing to the development and gradual breaking of finger-shaped structures. At higher forcing amplitudes F=2F = 2 and 3 for f/ω=0f/\omega= 0, 0.48, and 0.59, we observe the breakdown of finger-shaped structures in the first two oscillations of periodic forcing resulting in the early onset of turbulence. However, the turbulence is less intense and short-lived than at lower forcing amplitudes owing to the shorter sub-harmonic instability phase. This shorter sub-harmonic instability phase at higher FF is attributed to the higher growth rate of the fastest growing modes, which is 2.7 times larger for F=3F = 3 than F=1F = 1, as determined by linear stability analysis. An interesting finding is an increase in t.k.e.t.k.e. after an initial decrease for f/ω=0.59f/\omega= 0.59 due to the continuous triggering of the sub-harmonic instabilities.

We also analyze the energetics associated with the flow. An increase in t.k.e.t.k.e. for the cases with f/ω=0.48f/\omega= 0.48 at F=0.75,1F = 0.75, 1 implies an increase in the buoyancy flux BVB_V and therefore, APE also increases. Some part of this APE converts to BPE via irreversible mixing rate M\mathcal{M} related to the turbulent mixing, whereas the rest is converted back to t.k.e.t.k.e. via BVB_V. When the t.k.e.t.k.e. starts decreasing due to instability saturation, BVB_V also decreases, signifying a decrease in APE. Therefore, BPE saturates, and M\mathcal{M} approaches zero. Interestingly, for f/ω=0.48f/\omega= 0.48 the irreversible mixing sustains for a longer period than f/ω=0f/\omega= 0 due to a longer sub-harmonic instability phase. The cases with a higher rotation rate of f/ω=0.59f/\omega= 0.59 demonstrate a significant delay in the onset of sub-harmonic instability, owing to which BVB_V and APE remain zero. The initial concentration profile diffuses during this period to increase total PE at the rate ϕi\phi_i. This PE goes completely to BPE at the same rate since ϕd\phi_d becomes equal to ϕi\phi_i according to equation 3.4c. After the onset of the sub-harmonic instability, the t.k.e.t.k.e., BVB_V, and APE increase. However, these values remain significantly smaller than f/ω=0,0.48f/\omega= 0, 0.48 cases. In contrast to cases f/ω=0,0.48f/\omega= 0, 0.48, for f/ω=0.59f/\omega= 0.59 the bulk of the portion of APE expends to BPE via M\mathcal{M} resulting in a transfer of a small amount of APE back to t.k.e.t.k.e. reservoir. Therefore, in the subsequent oscillation, only a small t.k.e.t.k.e. is available to contribute to the external energy input from periodic forcing, signifying weaker turbulence compared to f/ω=0,0.48f/\omega= 0, 0.48 cases. Since the instabilities never saturate for f/ω=0.59f/\omega= 0.59, the t.k.e.t.k.e., BVB_V, and APE although small remain non-zero. The M\mathcal{M} remains greater than zero, signifying a continuous increase in BPE and irreversible turbulent mixing. At F=2,3F = 2, 3, M\mathcal{M} increases from the beginning of the vertical forcing for all f/ωf/\omega cases. However, the irreversible mixing for f/ω=0,0.48f/\omega= 0, 0.48 does not sustain for long owing to the quick saturation of the sub-harmonic instability. Since the instability never saturates for f/ω=0.59f/\omega= 0.59, M\mathcal{M} demonstrates continuous oscillations and therefore, continuous irreversible mixing. Finally, the mixing phenomenon is quantified using cumulative mixing efficiency. We conclude that for rotation rates (f/ω)2>0.25(f/\omega)^2 > 0.25 and at lower forcing amplitudes F=0.75,1F = 0.75, 1 the mixing process is more efficient because the major portion of APE expends to BPE via M\mathcal{M}, compared to (f/ω)2<0.25(f/\omega)^2 < 0.25.

Supplementary data. Supplementary movies are available at https://doi.org/**.*/jfm....

Acknowledgements. We gratefully acknowledge Dr. Vamsi K. Challamalla for providing us the computer program for the calculation of total PE, ϕ3∗\phi_3^*, BPE and ϕd\phi_d. We also want to thank the Office of Research and Development, Indian Institute of Technology Kanpur for the financial support through grant no. IITK/ME/2019/194. The support and the resources provided by PARAM Sanganak under the National Supercomputing Mission, Government of India at the Indian Institute of Technology Kanpur are gratefully acknowledged.

Declaration of interests. The authors report no conflict of interest.

Author contributions. The authors contributed equally to analysing data and reaching conclusions, and in writing the paper.

Appendix A

Floquet Analysis

The theorems related to the Floquet theory and the steps to solve the Mathieu equation (11) by using these theorems are briefly discussed here. Proofs of the theorems and solution steps can be found in [6 pp. 308–317].

Theorem 1 (Floquet’s theorem). Let dx⃗dt=A(t)x⃗\frac{\mathrm{d}\vec{x}}{\mathrm{d}t}=A(t)\vec{x}, be a first order linear differential system, where A(t)A(t) is TT-periodic matrix such that A(t+T)=A(t)A(t+T)=A(t), ∀t\forall t. This system has at

least one non-trivial solution x⃗=χ⃗(t)\vec{x}=\vec{\chi}(t) such that

χ⃗(t+T)=μχ⃗(t),∀ t(20)\vec{\chi}(t+T)=\mu\vec{\chi}(t),\quad\forall\ t \tag*{(20)}

where μ\mu is a characteristic number or Floquet multiplier.

If Φ(t)\Phi(t) is a fundamental solution matrix of system x⃗˙=A(t)x⃗\dot{\vec{x}}=A(t)\vec{x}, then Φ(t+T)\Phi(t+T) is also a fundamental matrix, and there exists a non-singular matrix CC such that,

Φ(t+T)=CΦ(t),∀ t(21)\Phi(t+T)=C\Phi(t),\quad\forall\ t \tag*{(21)}
C=Φ−1(t)Φ(t+T).(22)C=\Phi^{-1}(t)\Phi(t+T). \tag*{(22)}

Here, characteristic numbers, μ\mu's, are the eigenvalues of CC. We can obtain the matrix C=Φ(T)C=\Phi(T) for initial conditions at t=0t=0 such that Φ(0)=I\Phi(0)=I, where II is the identity matrix.

Theorem 2. For system x⃗˙=A(t)x⃗\dot{\vec{x}}=A(t)\vec{x}, where A(t)A(t) is TT-periodic, with characteristic numbers μ1,μ2,…μn\mu_{1},\mu_{2},\ldots\mu_{n}, the product of the characteristic numbers is obtained as:

μ1μ2⋯μn=exp⁡(∫0Ttr⁡{A(t)}).(23)\mu_{1}\mu_{2}\cdots\mu_{n}=\exp\left(\int_{0}^{T}\operatorname{tr}\{A(t)\}\right). \tag*{(23)}

The characteristic exponent (σ\sigma) of the system characterizing the growth rate of the instability is defined as eσT=μe^{\sigma T}=\mu.

Theorem 3. Let CC has nn distinct eigenvalues, μi\mu_i and corresponding σi\sigma_i, i=1,2,…,ni=1,2,\ldots,n. Then system x⃗˙=A(t)x⃗\dot{\vec{x}}=A(t)\vec{x} has nn linearly independent solutions of the form:

χ⃗i(t)=eσitP⃗i(t)(24)\vec{\chi}_{i}(t)=e^{\sigma_i t}\vec{P}_{i}(t) \tag*{(24)}

where P⃗i(t)\vec{P}_{i}(t) are the periodic vector functions with period TT.

Clearly, characteristic exponents eσite^{\sigma_i t}, will determine the behaviour of the solutions (A 4). Now, we apply above theorems to the mathieu equation (11). The equation (11) can be rewrite as:

a¨+(γ+α+βcos⁡τ)a=0,(25)\ddot{a}+(\gamma+\alpha+\beta\cos\tau)a=0, \tag*{(25)}

where,

γ=f2cos⁡2(θ)ω2,α=N2sin⁡2(θ)ω2,β=FN2sin⁡2(θ)ω2.(26)\gamma=\frac{f^{2}\cos^{2}(\theta)}{\omega^{2}},\quad\alpha=\frac{N^{2}\sin^{2}(\theta)}{\omega^{2}},\quad\beta=\frac{FN^{2}\sin^{2}(\theta)}{\omega^{2}}. \tag*{(26)}

We define a=Xa=X and a˙=Y\dot{a}=Y to express equation (A 5) as a first-order system, given as

\begin{bmatrix}\dot{X}\\\dot{Y}\end{bmatrix}= \begin{bmatrix}0&1\\-(\gamma+\alpha+\beta\cos\tau)&0\end{bmatrix} \begin{bmatrix}X\\Y\end{bmatrix}. \tag*{(27)}\end{bmatrix}

In the notation of Theorem 1,

A(τ)=[01−(γ+α+βcos⁡τ)0].(28)A(\tau)= \begin{bmatrix} 0&1\\ -(\gamma+\alpha+\beta\cos\tau)&0 \end{bmatrix}. \tag*{(28)}

Here, A(τ)A(\tau) is periodic with period 2π2\pi. As tr⁡{A(τ)}=0\operatorname{tr}\{A(\tau)\}=0, the product of the Characteristic numbers of A(τ)A(\tau) is μ1μ2=e0=1\mu_{1}\mu_{2}=e^{0}=1. For the following initial condition

Φ(0)=[1001],(29)\Phi(0)= \begin{bmatrix} 1&0\\ 0&1 \end{bmatrix}, \tag*{(29)}

we obtain the matrix C=Φ(T)C = \Phi(T) by numerically integrating (with MATLAB®) the system (A 7) from τ=0\tau= 0 to τ=2π\tau= 2\pi. μ1\mu_1 and μ2\mu_2 are the solutions of characteristic equation of CC

μ2−(sum of roots)μ+(product of roots)=0,(30)\mu^2 - (\text{sum of roots})\mu+ (\text{product of roots}) = 0, \tag*{(30)}

and using μ1μ2=1\mu_1\mu_2 = 1, we get

μ2−ϕ(γ,α,β)μ+1=0.(31)\mu^2 - \phi(\gamma,\alpha,\beta)\mu+ 1 = 0. \tag*{(31)}

Here, ϕ(γ,α,β)\phi(\gamma,\alpha,\beta) represents the sum of roots. The solutions μ1\mu_1 and μ2\mu_2 of the equation (A 11) are

μ1,μ2=12(ϕ±ϕ2−4).(32)\mu_1,\mu_2 = \frac{1}{2}\left(\phi\pm\sqrt{\phi^2 - 4}\right). \tag*{(32)}

Therefore, different values of ϕ\phi and corresponding μ1\mu_1, μ2\mu_2 (σ1\sigma_1 and σ2\sigma_2 will determine the behaviour of solutions. When ϕ>2\phi> 2: μ1\mu_1, μ2\mu_2 are both real and positive with μ1μ2=1\mu_1\mu_2 = 1, one of them, say μ1>1\mu_1 > 1 and μ2<1\mu_2 < 1. The corresponding characteristic exponents are real and have the form σ1=ξ>0\sigma_1 = \xi> 0, σ2=−ξ<0\sigma_2 = -\xi< 0 (because σ2=ln⁡(μ2)/T<0\sigma_2 = \ln(\mu_2)/T < 0). So, the general solution from Theorem 3 can be written as:

X(τ)=c1eξτP1(τ)+c2e−ξtP2(τ),(33)X(\tau) = c_1 e^{\xi\tau}P_1(\tau) + c_2 e^{-\xi t}P_2(\tau), \tag*{(33)}

where ξ>0\xi> 0, thus eξτe^{\xi\tau} corresponds to the exponential growth in time and solution becomes unstable with harmonic in nature. When ϕ<−2\phi< -2: μ1\mu_1, μ2\mu_2 are both real and negative with μ1μ2=1\mu_1\mu_2 = 1. The form of general solution is:

X(τ)=c1e(ξ+12i)τP1(τ)+c2e(−ξ+12i)τP2(τ).(34)X(\tau) = c_1 e^{(\xi+\frac{1}{2}i)\tau}P_1(\tau) + c_2 e^{(-\xi+\frac{1}{2}i)\tau}P_2(\tau). \tag*{(34)}

Here, eξτe^{\xi\tau} leads to the exponential growth of solution and eventually becomes unstable, with sub-harmonic response. When −2<ϕ<+2-2 < \phi< +2: μ1\mu_1, μ2\mu_2 are complex, we have σ1=iζ\sigma_1 = i\zeta and σ2=−iζ\sigma_2 = -i\zeta, where ζ\zeta is real, signifying that the solutions are bounded at all times and oscillatory but not necessarily periodic, and lies in stable parameter region. When ϕ=+2\phi= +2: μ1=μ2=1\mu_1 = \mu_2 = 1, one solution is periodic with period 2π2\pi and lies on the harmonic tongue of stability curve ϕ=2\phi= 2. When ϕ=−2\phi= -2: μ1=μ2=−1\mu_1 = \mu_2 = -1, one solution is periodic with period 4π4\pi and lies on the sub-harmonic tongue of stability curve ϕ=2\phi= 2.

References

  1. [2]Fernando, Hadrinja SS 1991 Turbulent mixing in stratified fluids. Annual review of fluid mechanics 23 (1), 455–493.
  2. [3]Fleury, M, Mory, M, Hopfinger, EJ & Auchere, D 1991 Effects of rotation on turbulent mixing across a density interface. Journal of Fluid Mechanics 223, 165–191.
  3. [5]Gregg, Michael C, D’Asaro, Eric A, Riley, James J & Kunze, Eric 2018 Mixing efficiency in the ocean. Annual review of marine science 10, 443–473.DOI
  4. [6]Jordan, Dominic & Smith, Peter 2007 Nonlinear ordinary differential equations: an introduction for scientists and engineers. OUP Oxford.
  5. [7]Naskar, Souvik & Pal, Anikesh 2022a Direct numerical simulations of optimal thermal convection in rotating plane layer dynamos. Journal of Fluid Mechanics 942.DOI
  6. [9]Nikurashin, Maxim, Vallis, Geoffrey K & Adcroft, Alistair 2013 Routes to energy dissipation for geostrophic flows in the southern ocean. Nature Geoscience 6 (1), 48–51.DOI
  7. [10]Noh, Yign & Long, Robert R 1990 Turbulent mixing in a rotating, stratified fluid. Geophysical & Astrophysical Fluid Dynamics 53 (3), 125–143.DOI
  8. [11]Pal, Anikesh 2020 Deep learning emulation of subgrid-scale processes in turbulent shear flows. Geophysical Research Letters 47 (12), e2020GL087005.DOI
  9. [12]Pal, Anikesh & Chalamalla, Vamsi K 2020 Evolution of plumes and turbulent dynamics in deep-ocean convection. Journal of Fluid Mechanics 889.DOI
  10. [13]Pal, Anikesh & Sarkar, Sutanu 2015 Effect of external turbulence on the evolution of a wake in stratified and unstratified environments. Journal of Fluid Mechanics 772, 361–385.DOI
  11. [14]Pal, Anikesh, de Stadler, Matthew B & Sarkar, Sutanu 2013 The spatial evolution of fluctuations in a self-propelled wake compared to a patch of turbulence. Physics of Fluids 25 (9), 095106.DOI
  12. [15]Peltier, WR & Caulfield, CP 2003 Mixing efficiency in stratified shear flows. Annual review of fluid mechanics 35 (1), 135–167.
  13. [16]Rosenberg, Duane, Pouquet, Annick, Marino, Raffaele & Mininni, Pablo Daniel 2015 Evidence for bolgiano-obukhov scaling in rotating stratified turbulence using high-resolution direct numerical simulations. Physics of Fluids 27 (5), 055105.DOI
  14. [17]Singh, Narinder & Pal, Anikesh 2022 The onset and saturation of the faraday instability in miscible fluids in a rotating environment. arXiv preprint arXiv:2207.13739.arxiv.org/abs/2207.13739
  15. [18]Talley, LD, Feely, RA, Sloyan, BM, Wanninkhof, R, Baringer, MO, Bullister, JL, Carlson, CA, Doney, SC, Fine, RA, Firing, E & others 2016 Changes in ocean heat, carbon content, and ventilation: a review of the first decade of go-ship global repeat hydrography. Annual review of marine science 8, 185–215.DOI
  16. [19]Winters, Kraig B & Barkan, Roy 2013 Available potential energy density for boussinesq fluid flow. Journal of Fluid Mechanics 714, 476–488.DOI
  17. [20]Winters, Kraig B, Lombard, Peter N, Riley, James J & D’Asaro, Eric A 1995 Available potential energy and mixing in density-stratified fluids. Journal of Fluid Mechanics 289, 115–128.
  18. [21]Wunsch, Carl & Ferrari, Raffaele 2004 Vertical mixing, energy, and the general circulation of the oceans. Annu. Rev. Fluid Mech. 36, 281–314.
  19. [22]Zoueshtiagh, F, Amiroudine, S & Narayanan, R 2009 Experimental and numerical study of miscible faraday instability. Journal of fluid mechanics 628, 43–55.DOI

Paper details

vlnc:W_yv36vqb3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 4

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents